Salt And Pepper Noise In Digital Image Filtering Algorithm Study

In the process of digital image formation, transmission and recording, as imaging systems, transmission medium and recording equipment are inadequate, so the images are polluted by a variety of noises. Among them, the maximum damage to the image quality is salt and pepper noise, because small amounts of noise will greatly damage the image details. In pattern recognition, computer vision, image analysis and video encoding and other fields, the image pre-processing is very important, because its result will directly affect the quality of following task. The purpose of image filter is to reduce noise, improve the signal to noise ratio, and protect the edge and detail information. In this respect, non-linear filter is more effective than the linear filter. Conventional median filter, as a kind of nonlinear filter based on order statistics, can effectively eliminate the salt and pepper noise and protect the edges and details in mages. Median filtering technique is widely used in image processing for its easy to realize and fast to calculate. In order to improve filtering performance, many improved median filtering algorithms are proposed.This paper presents the noise model and the foundational filtering knowledge, analyzes the researching development of salt and pepper noise filtering. The paper summarizes the main filtering methods such as switching filter, weighted median filter and the weighted average filter. Then the paper analyzes some important algorithms in details and compares their experiment results. Finally, the paper proposes three new algorithms: the salt and pepper noise filtering algorithm based on extremum and line detection, the adaptive salt and pepper noise removal algorithm based on histogram information, and Adaptive Iteration Filter for Removal of Random Value Impulse Noise. The salt and pepper noise filtering algorithm based on extremum and line detection can filter out salt and pepper noise while preserving black and white thin lines in image. The adaptive salt and pepper noise removal algorithm based on histogram information is better than most of conventional filtering algorithms for salt and pepper noise filtering of the idea model, because the filtering process does not need to select parameters and the algorithm has strong adaptability. It can restore the images which the noise rates are 0.8 and 0.9. The Adaptive Iteration Filter for Removal of Random Value Impulse Noise is more suitable for filtering random value of impulse noise, because it introduces the idea of switch filter according to random impulse noise characteristics and divides the algorithm into four steps: noise detection, noise filtering false detection pixel correction and iteration control. When random-value impulse noise being filtered, comparing with median filter from the experiments, the proposed algorithm protects image details well, needs no iterative parameters during filtering and has strong adaptive ability.